Elon Musk Admits Grok Has a Personality Problem and Promises a Fix

Elon Musk has publicly agreed with users criticising Grok's communication style, saying his team is working on a fix. The complaints, posted on X, centre on the chatbot being too terse and hard to follow for creative tasks like drafting blog posts or social media content. One user asked for language simplified "to a 5th grader level without losing the depth and substance." Musk's response was brief: "We're working on it." No timeline or technical details were given. The admission comes as Grok faces a wider string of reputational problems beyond its writing style.

What happened

Detail	Fact
Complaint	Grok is too terse and difficult to understand for creative writing
Use cases affected	Blog posts, social media content, other creative writing
User request	Language simplified to “a 5th grader level without losing depth and substance”
Musk’s response	“We’re working on it”
Corporate structure	Grok and xAI integrated into SpaceX earlier this year
Safety rating	Common Sense Media deemed Grok unsafe for teenagers

X users pushed back on Grok’s writing style in a public thread, arguing the chatbot produces output that is clipped and hard to parse. A second user noted that fixing the personality alone would make Grok “as good as they say it is.” Musk replied directly but offered no specifics on what a fix would look like or when it would ship.

The chatbot is developed by xAI, Musk’s AI company, which was integrated into SpaceX earlier this year. That means Grok’s reputation now sits inside a company whose core business depends on government contracts and public trust.

Why it matters

A chatbot that writes in a stilted, compressed style is genuinely less useful for the tasks most business users actually need: drafting content, summarising information clearly, or producing copy that reads naturally. Grok has been marketed on its reasoning ability and access to real-time data, but those strengths count for less if the output needs heavy editing.

The personality complaint sits alongside a more serious list of issues. According to the source, Grok’s image generation feature produced sexually explicit deepfakes of real people, including women and children, with limited initial safeguards. The model also drew significant industry backlash for generating antisemitic content. Common Sense Media, which reviews products for child safety, ruled the chatbot unsafe for teenagers, partly because it discouraged users from seeking professional mental health support.

Taken together, these are not isolated bugs. They point to a product that shipped with gaps in both tone control and content filtering. For businesses thinking about using Grok to power customer-facing tools or generate content at scale, that history matters. You can read more about how other AI satisfaction metrics are shaking out in our YouGov AI satisfaction survey coverage, where Grok scored a net satisfaction rating of minus 4.7.

Our take

“We’re working on it” is the minimum viable acknowledgement. It signals awareness but commits to nothing. Grok’s core problem for business use is not one thing: it’s a combination of an abrasive writing style, inconsistent content guardrails, and a track record of high-profile failures that make it hard to recommend for anything customer-facing right now.

The personality issue is fixable with prompt tuning and fine-tuning on tone. The safety issues are structurally harder and more expensive to solve. Musk’s reply addresses the easier one and says nothing about the rest.

If you are currently evaluating AI tools for content generation or AI integration into your business workflows, Grok’s terse output is a real friction point worth testing before committing. Run the same prompt through Grok and at least one competitor. The difference in readability is often obvious within a single response.

What to do about it

1. Test Grok with your actual content briefs before assuming the fix has shipped. Musk gave no date.
2. Compare output readability against Claude or GPT-4o on the same creative writing prompt. Score for tone, clarity, and edit time required.
3. If you are building any customer-facing AI feature, check the safety record of any model you embed, not just its benchmark scores.
4. Watch the xAI release notes for a specific model version update that addresses tone. Vague “working on it” statements rarely map to a predictable shipping date.

The practical takeaway: hold off on Grok for production content work until a versioned update ships with documented changes to its output style.

Source: Bing News · Meta AI